Fix LoadBalancer tests to be provider-agnostic

[danwinship]

What this PR does / why we need it:

Drops all the provider checks from the LoadBalancer tests and instead just marks them all [Feature:LoadBalancer]. We will need corresponding updates in test-infra...

At the moment it doesn't make any attempt to distinguish between what different LoadBalancer-supporting platforms support (#123714). People who want to run them will have to just manually --ginkgo.skip the tests they can't pass for now.

(Note: originally this PR was about trying to get the LB tests to pass under cloud-provider-kind, but now it is just about re-enabling them on GCE.)

Which issue(s) this PR fixes:

Fixes #124338

Does this PR introduce a user-facing change?

NONE

/kind cleanup
/sig network
/priority important-soon
/assign @aojea

[k8s-ci-robot]

This issue is currently awaiting triage.

If a SIG or subproject determines this is a relevant issue, they will accept it by applying the triage/accepted label and provide further guidance.

The triage/accepted label can be added by org members by writing /triage accepted in a comment.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[danwinship]

pull-kubernetes-e2e-kind-cloud-provider-loadbalancer results:

 Summarizing 7 Failures:
  [FAIL] [sig-network] LoadBalancers [Feature:LoadBalancer] [It] should be able to preserve UDP traffic when server pod cycles for a LoadBalancer service on the same nodes [sig-network, Feature:LoadBalancer]
  k8s.io/kubernetes/test/e2e/network/loadbalancer.go:866
  [FAIL] [sig-network] LoadBalancers [Feature:LoadBalancer] [It] should be able to change the type and ports of a UDP service [Slow] [sig-network, Feature:LoadBalancer, Slow]
  k8s.io/kubernetes/test/e2e/network/service.go:591
  [FAIL] [sig-network] LoadBalancers ExternalTrafficPolicy:Local [Feature:LoadBalancer] [Slow] [It] should work for type=LoadBalancer [sig-network, Feature:LoadBalancer, Slow]
  k8s.io/kubernetes/test/e2e/network/loadbalancer.go:990
  [FAIL] [sig-network] LoadBalancers [Feature:LoadBalancer] [It] should be able to preserve UDP traffic when server pod cycles for a LoadBalancer service on different nodes [sig-network, Feature:LoadBalancer]
  k8s.io/kubernetes/test/e2e/network/loadbalancer.go:736
  [FAIL] [sig-network] LoadBalancers [Feature:LoadBalancer] [It] should have session affinity work for LoadBalancer service with Cluster traffic policy [Slow] [LinuxOnly] [sig-network, Feature:LoadBalancer, Slow]
  k8s.io/kubernetes/test/e2e/network/service.go:266
  [FAIL] [sig-network] LoadBalancers [Feature:LoadBalancer] [It] should be able to switch session affinity for LoadBalancer service with Local traffic policy [Slow] [LinuxOnly] [sig-network, Feature:LoadBalancer, Slow]
  k8s.io/kubernetes/test/e2e/network/service.go:266
  [FAIL] [sig-network] LoadBalancers [Feature:LoadBalancer] [It] should be able to switch session affinity for LoadBalancer service with Cluster traffic policy [Slow] [LinuxOnly] [sig-network, Feature:LoadBalancer, Slow]
  k8s.io/kubernetes/test/e2e/network/service.go:266 

So all of the UDP LoadBalancer tests failed (which is a known bug, kubernetes-sigs/cloud-provider-kind#58), and most of the SessionAffinity tests. ("should have session affinity work for LoadBalancer service with Local traffic policy" passed, but none of the others did.)

[danwinship]

/test pull-kubernetes-e2e-kind-cloud-provider-loadbalancer
seeing if failures are consistent or flakes...

[aojea]

/test pull-kubernetes-e2e-kind-cloud-provider-loadbalancer

[aojea]

great, we moved from

Kubernetes e2e suite: [It] [sig-network] LoadBalancers [Feature:LoadBalancer] should be able to preserve UDP traffic when server pod cycles for a LoadBalancer service on different nodes expand_more	3m18s
Kubernetes e2e suite: [It] [sig-network] LoadBalancers [Feature:LoadBalancer] should have session affinity work for LoadBalancer service with Cluster traffic policy [Slow] [LinuxOnly] expand_more	10m47s
Kubernetes e2e suite: [It] [sig-network] LoadBalancers ExternalTrafficPolicy:Local [Feature:LoadBalancer] [Slow] should work for type=LoadBalancer expand_more	1m20s
Kubernetes e2e suite: [It] [sig-network] LoadBalancers [Feature:LoadBalancer] should be able to preserve UDP traffic when server pod cycles for a LoadBalancer service on the same nodes expand_more	2m12s
Kubernetes e2e suite: [It] [sig-network] LoadBalancers [Feature:LoadBalancer] should be able to switch session affinity for LoadBalancer service with Local traffic policy [Slow] [LinuxOnly] expand_more	10m13s
Kubernetes e2e suite: [It] [sig-network] LoadBalancers [Feature:LoadBalancer] should be able to change the type and ports of a UDP service [Slow] expand_more	3m46s
Kubernetes e2e suite: [It] [sig-network] LoadBalancers [Feature:LoadBalancer] should be able to switch session affinity for LoadBalancer service with Cluster traffic policy [Slow] [LinuxOnly] expand_more

To only

Kubernetes e2e suite: [It] [sig-network] LoadBalancers [Feature:LoadBalancer] should be able to change the type and ports of a UDP service [Slow] expand_more	16m19s
Kubernetes e2e suite: [It] [sig-network] LoadBalancers ExternalTrafficPolicy:Local [Feature:LoadBalancer] [Slow] should work for type=LoadBalancer expand_more

This will not pass as the loadbalancer is type proxy, so it will always replace the source ip, we may try to use something like this https://www.envoyproxy.io/docs/envoy/latest/intro/arch_overview/other_features/ip_transparency#original-source-listener-filter but it may not work in all environments

Kubernetes e2e suite: [It] [sig-network] LoadBalancers ExternalTrafficPolicy:Local [Feature:LoadBalancer] [Slow] should work for type=LoadBalancer expand_more

The UDP one fails because is waiting for a rejection and we have discussed the same for TCP, is not clear we can impose this behavior for loadbalancers in general)

kubernetes/test/e2e/network/service.go

Lines 609 to 621 in 4878acc

	// testRejectedUDP tests that the given host rejects a UDP request on the given port.
	func testRejectedUDP(host string, port int, timeout time.Duration) {
	pollfn := func() (bool, error) {
	result := pokeUDP(host, port, "echo hello", &UDPPokeParams{Timeout: 3 * time.Second})
	if result.Status == UDPRefused {
	return true, nil
	}
	return false, nil // caller can retry
	}
	if err := wait.PollImmediate(framework.Poll, timeout, pollfn); err != nil {
	framework.Failf("UDP service %v:%v not rejected: %v", host, port, err)
	}
	}

[FAILED] UDP service 172.18.0.6:81 not rejected: timed out waiting for the condition

[aojea]

with

diff --git a/test/e2e/network/loadbalancer.go b/test/e2e/network/loadbalancer.go
index e68e9153d11..08896bbecae 100644
--- a/test/e2e/network/loadbalancer.go
+++ b/test/e2e/network/loadbalancer.go
@@ -495,8 +495,8 @@ var _ = common.SIGDescribe("LoadBalancers", ginkgo.Serial, func() {
                err = udpJig.Scale(ctx, 0)
                framework.ExpectNoError(err)
 
-               ginkgo.By("looking for ICMP REJECT on the UDP service's LoadBalancer")
-               testRejectedUDP(udpIngressIP, svcPort, loadBalancerCreateTimeout)
+               ginkgo.By("UDP service's LoadBalancer should not be reachable")
+               testNotReachableUDP(udpIngressIP, svcPort, loadBalancerCreateTimeout)
 
                ginkgo.By("Scaling the pods to 1")
                err = udpJig.Scale(ctx, 1)

for the other test should work for type=LoadBalancer we should add a new Label as Feature:LoadbalancerL4 or Layer4 or something that indicates that the loadbalancer provisioned operates at L4

[danwinship]

This will not pass as the loadbalancer is type proxy

The e2e test should check the LoadBalancerStatus and belatedly skip the test in that case. (If either ingress.IPMode == v1.LoadBalancerIPModeProxy or ingress.IP == "")

[aojea]

great , those two tests that are failing now will be fixed once we make them aware of the proxy mode as Dan says in previous comment and we implement that in cloud-provider-kind kubernetes-sigs/cloud-provider-kind#63

Kubernetes e2e suite: [It] [sig-network] LoadBalancers ExternalTrafficPolicy:Local [Feature:LoadBalancer] [Slow] should work for type=LoadBalancer expand_more 1m32s
Kubernetes e2e suite: [It] [sig-network] LoadBalancers ExternalTrafficPolicy:Local [Feature:LoadBalancer] [Slow] should handle updates to ExternalTrafficPolicy field expand_more

[dims]

@danwinship please re-add the entire test/e2e/network/loadbalancer.go back in this PR once #124519 lands. thanks!

[danwinship]

(FTR I'm doing this now just to keep all (most?) of the renaming in one PR; I didn't want people updating their CI configs to skip all the "ESIPP" tests and then have us rename them later...)

[aojea]

@danwinship linter error

{ ScriptError 
test/e2e/network/loadbalancer.go:1453:6: ST1003: var previousHttpErrors should be previousHTTPErrors (stylecheck)
	var previousHttpErrors uint64 = 0
	    ^
test/e2e/network/loadbalancer.go:1500:3: ST1003: var currentHttpErrors should be currentHTTPErrors (stylecheck)
		currentHttpErrors := atomic.LoadUint64(&httpErrors)
		^
test/e2e/network/loadbalancer.go:1504:3: ST1003: var partialHttpErrors should be partialHTTPErrors (stylecheck)
		partialHttpErrors := currentHttpErrors - previousHttpErrors
		^

[k8s-ci-robot]

@danwinship: The following test failed, say /retest to rerun all failed tests or /retest-required to rerun all mandatory failed tests:

Test name	Commit	Details	Required	Rerun command
pull-kubernetes-e2e-kind-cloud-provider-loadbalancer	67e0c51	link	false	/test pull-kubernetes-e2e-kind-cloud-provider-loadbalancer

Full PR test history. Your PR dashboard. Please help us cut down on flakes by linking to an open issue when you hit one in your PR.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here.

[aojea]

/test pull-kubernetes-e2e-gce-100-performance

[danwinship]

/retest-required

[aojea]

/test pull-kubernetes-e2e-gce-100-performance

[danwinship]

@aojea what are you running the perf test for?
anyway, I think this is ready to merge?

[aojea]

@aojea what are you running the perf test for?

wrong ones sorry, I wanted to check kubernetes/test-infra#32585

/lgtm
/approve

Let's iterate

[k8s-ci-robot]

LGTM label has been added.

Git tree hash: a8e87178b1c683203d68ba3d524f8f3984407eea

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: aojea, danwinship

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• test/e2e/feature/OWNERS [aojea,danwinship]
• test/e2e/network/OWNERS [aojea,danwinship]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment